A quick note today. Nvidia reported its latest results. We have worked in semis a long time, and we remember in the distant past, say four years ago, when only a few hundred people cared much about Nvidia’s results (outside of its employees). Now it sometimes seems like the entire fate of Western Civilization depends on their results, such is the general frenzy around AI. At least, the results were very good.

The company reported revenue and EPS ahead of expectations, $22 billion on consensus of $20 billion and $5.16 on $4.63, respectively. The company also guided to revenue of $22 billion ahead of expectations of $24 billion. These are solid results, albeit not quite as awe-inspiring as the company’s last three sets of earnings. To put this in context, Nvidia’s total revenue in 4Q of last year was $6 billion, growing to $22 billion in the latest quarter.

We are not going to dig into all their commentary and numbers, but a few things stood out for us. Most notable was the company’s commentary that 40% of their data center revenue are from products being used for Inference workloads. AI inference is the mostly hotly contested sub-segment of the market now, one upon which most of Nvidia’s competitors have built their hopes. The inference build out is still in early days, so there is no guarantee that Nvidia can maintain this level. That being said, there is no reason to assume it cannot go up even further.

We also think it is important to look at Nvidia’s results in the context of broader changes in the market for data center processors. Nvidia now accounts for 75% of that market, up from 73% last quarter and 10% in 2019. That market has grown from $6 billion to $25 billion in just over 4 years. So they are making massive gains in a market that is growing at 36% at compound annual growth rate (CAGR) over that period. Looked at another way, if we exclude Nvidia, the market has only grown at a 3% rate over that period, a time which included major hyperscaler build outs. Over that period, Nvidia’s data center revenue grew at a 114% CAGR, it has been been more than doubling every year for almost five years.

With growth like that, for a company of this size, it is easy to see how they have attracted so much interest. Of course, there are two downsides to this. First, retail participation in the stock seems to be very high, and this is not a great stock for retail investors to be able to track. Secondly, the company is now priced for perfection, with considerably heightened expectations across the Street. This meant that in the run-up to this week’s results, online discourse about the company bears all the worst hallmarks of Internet commentary, with more emotion than seems good for anyone. This is dangerous for a company in a highly cyclical industry with a history of highly cyclical results. Recall that Nvidia has a track record of badly missing expectations every few years. One reader told us that by his count, in Nvidia’s 25 years as a publicly traded company it has seen three drawdowns of over 50%. As much as we believe there are fundamental pillars to Nvidia’s continued growth, we are not quite convinced that the market views the stock rationally. Nvidia is an important company, and it is doing well. How much better its stock will do is a totally different question.